RECUPERACIÓN DE CALOR DE RUEDA DESHUMECTADORA EN HORNO

The contrasting results in this chapter with those in Chetty and Hendren (2018a) suggest that childhood environment has become more important for upward mobil- ity over time. In terms of the framework described above, βChildhood _{has increased}

relative to βLabor. Places that are more conducive for a child’s development today are thus likely to exhibit higher levels of upward mobility. One metric of the con- duciveness of a place is the proportion of children who grew up there and attended college by their late teens or young adult years.154 For the contemporary cohorts in Chetty et al. (2014), these rates tend to be higher among those who were raised in

153_{For simplicity, this framework assumes away internal migration – individuals reside in the same}

place as children and as adults. Hence the use of a common subscript c for both the childhood and labor market characteristics of a place when aggregating the two effects. How reasonable this simplifying assumption is depends on the proportion of individuals who remain in their childhood CZs into adulthood.

the non-industrial Midwest.155 _{The growing importance of childhood environment}

could therefore have improved upward mobility in this region and explain part of the non-industrial Midwest’s transformation into the land of opportunity.156 This begs the question: why did childhood environment become more important for economic outcomes and upward mobility? I propose that the rising importance and levels of human capital may have played a role.157

1.7.1.1 The Rising Importance of Human Capital

The connection between childhood environment and upward mobility may have grown stronger because human capital became more important for success in the labor mar- ket. Specifically, access to top jobs has become increasingly dependent on an indi- vidual’s education level. To illustrate this, I regress an indicator for persons who completed at least 4 years of college on a dummy for top jobs, controlling for a quadratic in age.158 _{The coefficient on top jobs captures the age-adjusted difference}

in the share of college workers between top jobs and lower-ranked occupations. Us- ing lower-ranked occupations as a comparison group is necessary because the level of

155_{Chetty et al.’s (2014) data pool both genders, all races, and all nativities together.}

156_{This implication depends on where good childhood environments are located in the present.}

The corresponding geographic distribution in the early 20th century, on the other hand, is less rele- vant here because childhood environment was unlikely to have been important for upward mobility historically, as shown earlier.

157_{Recent work by Rothstein (2018) argues that human capital accumulation accounts for a small}

portion of the geographic variation in upward mobility observed in Chetty et al.’s (2014) data – this appears to be at odds with my proposed mechanism. Two points are worth noting here. First, Rothstein (2018) states that his “analysis is purely observational”. Caution is thus necessary when interpreting his results. Second, Rothstein (2018) finds that two-fifths of the geographic variation in upward mobility is due to differences in spousal earnings and non-labor income, both of which are included in Chetty et al.’s (2014) baseline measure of upward mobility. However, Chetty et al. (2014) show that regardless of whether mobility is computed using household income, individual income, or individual earnings (excludes capital and other non-labor income), the map of upward mobility remains almost unchanged (Online Appendix Table VII in their paper). This cannot be the case if two-fifths of the variation in the initial measure based on household income had been due to differences in spousal earnings and non-labor income, as Rothstein (2018) suggests. Put differently, Rothstein’s (2018) conclusions are inconsistent with Chetty et al.’s (2014).

158_{The population used here comprises white men aged 25-54, who live in households, and who}

education attainment has risen nationally over the 20th century, as shown in Figure 1.20. This trend has to be taken into account in order to accurately determine how the relation between human capital and top jobs has evolved. I estimate the afore- mentioned regression separately for each census year from 1940 to 2015 and plot the resulting coefficients in Figure 1.21.159 _{I find that the composition of workers in top}

jobs has become increasingly skilled relative to that in lower-ranked occupations, with the divergence being particularly rapid during the middle of the 20th century. Such a development would make it more difficult for persons without a college education to gain access to top occupations. Top jobs may still be able to facilitate upward mo- bility but only if one has the necessary education to work in these professions. Since human capital is cumulative, how much education an individual can acquire depends on how good his or her childhood environment was at developing children’s cognitive abilities. The link between childhood environment and adult economic outcomes will thus strengthen as human capital becomes more important in the labor market.

1.7.1.2 The Rising Levels of Human Capital

In addition to its increasing importance in the labor market, the rising levels of human capital per se depicted in Figure 1.20 may have also amplified the role of childhood environment. The intuition is as follows. Suppose as before that neighborhoods have a causal effect on children’s cognitive development, thus influencing how far they can go up the education ladder. During the early 20th century, where a person grew up in was less important for education attainment because average schooling levels were low to begin with. Even if one was raised in a better environment, it still may not have been worthwhile acquiring a lot of human capital. Today, with the average level of schooling being much higher, childhood environment could have a

159_{The thresholds used to define top jobs are computed separately for each census year, based}

on income scores for 1940 and based on the median total income of working white men in a given occupation for all other years.

greater part in education production. This is because human capital is cumulative in nature, which implies that those with an early head start will have an advantage in human capital accumulation later in life. Consequently, a stronger relation between childhood environment and education attainment will be observed in the present. This growing connection will also be reflected in income and income ranks due to the positive return to schooling. To illustrate the mechanics described here more formally, I use an extension of Card’s (2001) endogenous schooling model.

Consider an infinitely-lived individual who is deciding how many years of schooling S to acquire in order to maximize the present value of his or her lifetime earnings:160

max S Z ∞ S wH(S)e−rt· dt ⇒ max S w rH(S)e −rS (25)

where w is the compensation per unit of human capital and r is the discount rate. H(.) is the human capital production function which transforms S into effective units of human capital that are valued in the labor market, with H0(.) > 0 and H00(.) < 0. This setup leads to the usual first order condition H_H(S)0(S) = r that pins down the optimal years of schooling S∗.

Moving away from the unconstrained model, whether an individual can reach S∗ could depend on where he or she was raised. To formalize this, I introduce the impact of childhood environment as a constraint on the years of schooling a person can potentially complete:

S ∈ [0, g(x(nc))] (26)

The variable x(.) is the quantity of positive exposure units that an individual accu- mulates while growing up in a particular neighborhood, where CZs are once again used to define neighborhoods. It depends on the quality of the neighborhood one

160_{One should think of this decision as being made when the individual is in his or her late teens or}

was raised in, denoted by nc. A higher nc indicates a better childhood environment.

Thus, xnc > 0. This setup implicitly assumes that there is no internal migration dur-

ing childhood, though movements as adults are allowed. All individuals thus spend their entire childhoods in one locality, which is why x(.) varies only with the place where a person was raised and not the amount of time spent there. Once x(.) has been determined, the function g(.) then maps the positive exposure units to the max- imum attainable years of schooling. This mapping could reflect a combination of the causal effect of neighborhoods on children’s cognitive development and the cumula- tive nature of human capital. To simplify the discussion below, I assume that the maximum attainable level of schooling varies with neighborhood quality at a constant rate: g0(.)xnc = γ > 0. That is, growing up in a place that is one unit better relaxes

the schooling constraint by γ.

To make the connection between the rising levels of human capital and the increas- ing importance of childhood environment, consider a simple counterfactual. Suppose that individuals who were originally raised in CZ c are now given an alternative life where they are raised in a CZ that is 1 unit better. This relaxes constraint (26) by γ. How will each person adjust his or her level of education in response to the change in setting? This depends on the extent to which the schooling constraint was binding in the original CZ, formally expressed as S_i∗ − g(x(nc)) where the optimal level of

schooling is allowed to vary across individuals.161,162 _{Three types of responses may be}

161_{One way of introducing heterogeneity in S}∗ _{is to allow the discount rate r to vary by person.} 162_{The implicit assumption here is that the change in setting does not alter the optimal level of}

observed:             

Case 1: S_i∗− g(x(nc)) ≤ 0, Increase in schooling: 0

Case 2: 0 < S_i∗− g(x(nc)) < γ, Increase in schooling: (0, γ)

Case 3: S_i∗− g(x(nc)) ≥ γ, Increase in schooling: γ

Those in Case 1 were not constrained in the initial setting. The years of schooling they complete will thus remain unchanged even in a better environment. Individuals in Case 2 will experience a relaxation of the schooling constraint and increase the amount of human capital acquired by S_i∗ − g(x(nc)). Case 3 refers to persons for

whom the constraint was most binding originally. They will raise their schooling levels by the full extent to which the constraint is relaxed: γ.

Because the counterfactual is based on a unit change in neighborhood quality, the marginal effect of a better childhood environment can be computed by averaging the individual responses: 1 Q Q X i=1 Responsei = 1 Q Q0 X i=1 (S_i∗− g(x(nc))) + 1 Q Q00 X i=1 γ = Q 0 Q  1 Q0 Q0 X i=1 S_i∗− g(x(nc))  +Q 00 Qγ ≤ γ (27)

where Q is the total number of individuals, Q0 refers the subset of persons in Case 2, and Q00denotes those in Case 3. Equation (27) reveals how the rise in education levels can amplify the estimated importance of childhood environment. All else equal, the growth in human capital documented in Figure 1.20 suggests that both the proportion of persons facing a binding constraint (Q_Q0 and Q_Q00) and the mean optimal level of schooling for individuals in Case 2 (_Q10

PQ0

i=1S ∗

than in the past. One should thus expect to find stronger childhood exposure effects on education attainment today. Coupled with the positive return to schooling in both the historical and contemporary periods (Card 2001; Clay et al. 2016), the stronger exposure effects in the present will also be reflected in income ranks. This replicates how childhood environment has grown more important for upward mobility over time. 1.7.1.3 Implication of the Increasing Importance and Levels of

Human Capital

The increasing relevance of childhood environment for later-life outcomes either through the rising importance or levels of human capital suggests a testable implication: the ability of neighborhoods to produce human capital should be more strongly associ- ated with upward mobility today than historically. To check this, I use the share of individuals with more than 8 years of schooling to proxy for the quality of neighbor- hoods during the early 20th century, while college attendance rates are used as the corresponding metric for the contemporary period.163 _{Both measures are based on}

where a person grew up in. The former has a simple correlation of 0.468 with the historical geography of upward mobility, while the latter has a higher correlation of 0.669 with the contemporary landscape of upward mobility, thus offering some sup- port for the mechanisms proposed here.164,165 The stronger correlation observed in the present suggests that the land of opportunity has shifted in favor of places that are more conducive for a child’s development.

1.7.1.4 A Role for Changes in Internal Migration?

163_{The former is based on the 1910-1940 linked sample, while the latter is taken from CFHJP. Both}

refer to white sons from below-median income households. I use 8 years of schooling as the metric for the historical setting as this provides the closest comparison with CFHJP’s college attendance rate in terms of national shares. 39.7 percent of individuals in the historical sample have more than 8 years of schooling while the weighted college attendance rate in CFHJP is 47.7 percent.

164_{This is based on 627 common CZs.}

165_{The difference in magnitudes is not driven primarily by the South. Restricting the sample to}

Apart from the changes in human capital, could changes in the degree of residential mobility have reduced the importance of local labor markets for upward mobility, thus indirectly amplifying the role of childhood environment? Molloy et al. (2011) document that lifetime interstate migration rates have increased over the course of the 20th century. This suggests that individuals are less bound by the labor markets they started out in. Of greater importance then would be the amount of human capital they carry to their destinations. As a result, local labor markets may become less important relative to childhood environment, even if the latter has not actually grown in absolute importance. Given that the non-industrial Midwest has always been a leader in human capital production, both historically during the high school movement (Goldin and Katz 2008) and in more recent times (Chetty et al. 2014), the increase in internal migration would then enable Midwesterners to make better use of their human capital advantage. This, in turn, might have contributed to the non-industrial Midwest’s rise as the land of opportunity.

While the national rise in internal migration rates may diminish the relevance of local labor markets, I argue that it cannot by itself explain why the non-industrial Midwest overtook the coastal regions and the industrial Midwest in terms of upward mobility. Figure 1.22 illustrates why: even though migration within the country has increased on average, this is not the case for those from the non-industrial Midwest. The dotted line with square markers plots the share of 30-39 year old native-born white men living outside of their state of birth in each census year from 1900 to 2015.166 _{There is a general upward trend in interstate migration rates from 1940}

till 1980, after which it flattens out. This is broadly consistent with Molloy et al. (2011).167 Focusing on those who were born in the non-industrial Midwest, however,

166_{The 30-39 age range allows for a time series with non-overlapping cohorts.}

167_{One difference is that Molloy et al. (2011) find a modest upward trend from 1900 to 1930 as}

well, in contrast to the decline in Figure 1.22. Note that Molloy et al. (2011) do not impose race or gender restrictions on their sample.

paints a different picture, as depicted by the dashed line with triangle markers. If anything, 30-39 year olds in 2015 – who overlap with Chetty et al.’s (2018) cohorts – are even less likely to move across states than the cohorts from the first half of the 20th century. A similar pattern is observed when the population is further restricted to individuals who were born in the non-industrial Midwest and who also moved out of the region, as illustrated by the solid line with circle markers.168 Thus, persons born in the non-industrial Midwest have not necessarily become more likely to move to different parts of the country.169 _{The national rise in internal migration on its own}

is thus unlikely to explain the relative improvement in upward mobility for persons raised in the non-industrial Midwest.

1.7.2 Changes in the Spatial Distribution of the Determinants of Mobility In addition to how the determinants of upward mobility have changed, changes in the spatial distribution of these determinants may have also contributed to the shift in the land of opportunity. To shed light on this, I consider three place-specific characteristics, each of which is associated with either the childhood environment or labor market structure of a place: income segregation, the extent of industrialization or deindustrialization, and the degree of competition from immigrants.170 I ask if the geographic patterns of these features have evolved in a way that is consistent with the reversal in the landscape of upward mobility.

1.7.2.1 Income Segregation

168_{For this subsample, there is a more pronounced hump in the middle of the 20th century compared}

with the earlier population of all persons born in the non-industrial Midwest.

169_{Figure A.21 in Appendix A.3 shows that individuals born in the non-industrial Midwest are}

increasingly likely to move to labor markets that are more human capital intensive. However, these migration rates do not exceed 2 percent and are thus unlikely to play an important role in explaining the changing landscape of upward mobility.

170_{One feature that is not discussed here is the racial composition of a place. Derenoncourt (2018)}

shows that the second wave of the Great Migration, which altered the racial composition of northern cities dramatically, did not affect upward mobility for white men.

Given the widespread transformation of America’s residential landscape over the course of the 20th century, the spatial patterns of income segregation may have changed as well. This could alter the geography of upward mobility if income segrega- tion has consistently been harmful for mobility.171 _{To investigate this possibility, one}

needs to construct historical and contemporary estimates of income segregation for each CZ. Given the focus on whites in this chapter, these metrics should go beyond